Natural Language-Based Management of Computing Resources Executing Radio Access Network Workloads

This application is a continuation of U.S. patent application Ser. No. 18/335,745, filed Jun. 15, 2023, the content of which application is hereby expressly incorporated herein by reference in its entirety.

As cloud computing rapidly gains popularity, more and more data and/or services are stored and/or provided online via network connections. Providing an optimal and reliable user experience is an important aspect for cloud service providers that offer network services. In many scenarios, a cloud service provider may provide a service to thousands or millions of users (e.g., customers, clients, etc.) geographically dispersed around a country, or even the world. In order to provide this service, a cloud service provider often utilizes different resources, such as server farms, hosted in various datacenters. Access to these resources is typically provided by a cloud platform which operates the datacenters. In addition, the service can be constructed of various software components such as virtual machines, containers, and requisite management infrastructure. These software components may be collectively referred to as a cluster.

In recent years, a particular application space that has experienced the significant impact of cloud computing is radio access networks (RAN). Generally described, a radio access network is a component of a mobile telecommunication system that connects various devices (e.g., mobile phones, computers) to a core network (e.g., 5G, 4G LTE). Traditional radio access networks typically comprise many stand-alone base stations where each base station provides service to devices within a local geographical area. In addition, each base station possesses an individual set of resources (e.g., computing, cooling, power) to enable the base station to process and transmit its own signal to and from devices and forwards data payloads to the core network. Hence, the “cellular” nature of a cellular network.

There are many well-known limitations of these traditional network architectures. Most prominently, the isolated nature of the base stations can give rise to subsequent drawbacks. For instance, due to limited availability in the frequency spectrum, different base stations oftentimes utilize the same frequencies which can lead to interference between base stations. This issue can be exacerbated when a network operator adds additional base stations to the network to increase capacity. In another example, base stations can be highly resource inefficient. Due to the mobile nature of network users, traffic at a given base station can fluctuate dramatically. However, average utilization across all base stations of a network can often be very low, only intermittently experiencing spikes in traffic. In addition, traditional base stations often lack the ability to share computing resources with other base stations. As such, individual base stations may typically be designed for worst-case scenario processing loads thereby leading to poor resource efficiency and increased operating costs.

In contrast, cloud radio access networks (C-RANs), which can also be referred to as virtualized radio access networks (V-RAN) can leverage the computing power and flexibility of cloud platforms to virtualize radio access network functions. Consequently, cloud radio access networks can address many of the technical challenges facing traditional base station style radio access networks. For example, virtualizing functions that were previously performed by discrete computing devices enables the cloud radio access network to scale up and scale down available resources based on network conditions (e.g., traffic). In another example, centralizing network resources can streamline management and improve reliability. However, many existing tools for managing and orchestrating cloud computing resources such as Kubernetes may not have been designed with radio access network workloads in mind. As such, managing a cloud radio access network can be a highly complex task often requiring extensive manual customization. It is with respect to these and other considerations that the disclosure made herein is presented.

The techniques disclosed herein enhance computing systems that execute radio access network (RAN) workloads through a natural language interface for managing computing resources. As mentioned above, radio access networks are components of a telecommunications system that connect devices such as mobile phones to the broader core network. While traditional systems utilized physical base stations to implement a radio access network, recent developments have seen rapid virtualization of radio access network functions using cloud computing infrastructure (e.g., a datacenter). However, managing computing resources for a cloud radio access network (C-RAN) can be a deeply complex task as many orchestration tools such as Kubernetes may not account for specific needs of radio access network workloads. For example, many default components can be ideal for standard web-based workloads in which computing resources can be freely enabled (e.g., scaled) and/or disabled (e.g., killed). In contrast, radio access network workloads must consider state when managing computing resources as freely enabling and/or disabling resources can degrade service quality (e.g., dropped calls).

To address the technical challenges of utilizing a computing system to implement specialized workloads such as a radio access networks, many operators develop custom components such as custom resource definitions, custom controllers, and custom workloads for configuring the radio access network. Developing such custom components can be a demanding technical challenge with a high risk of errors and complexity for end users (e.g., network technicians). Moreover, these custom components can require continuous work to ensure compatibility with standard protocol specifications. In still another technical challenge, many existing tools can lack support for important functionality such as migrating the radio access network without disruptions. As such, within the computing technical paradigm, operators must contend with highly complex processes for managing computing resources and/or workloads while maintaining consistent service quality.

To address these and other technical challenges, the system discussed herein provides a natural language interface to enable a user (e.g., a system engineer, a technician) to manage computing systems and radio access networks utilizing a natural language input (e.g., English). In various examples, the natural language input can define a desired outcome for the computing system. For instance, the natural language input can be “node A needs to be upgraded” in which the desired outcome is an upgrade to a particular computing resource.

A natural language processing module of the disclosed system can accordingly analyze the natural language input to generate a custom contextual query (e.g., a prompt). In some examples, the query can be the natural language input alone to instruct a large language model to execute the task defined by the natural language input. In other examples, the disclosed system can retrieve auxiliary information to supplement the natural language input. In this way, the natural language processing function can produce a more robust query to provide additional context to the large language model and improve performance of the large language model.

The custom contextual query can be subsequently provided to the large language model for execution. In various examples, the custom contextual query can cause the large language model to generate an operation sequence comprising a plurality of commands performing the necessary operations on the computing system and achieve the desired outcome defined by the natural language input. In a specific example, consider again the natural language input specifying that “node A needs to be upgraded.” In response, to this natural language input, the large language model can generate an operation sequence that prevents new radio access network tasks from being scheduled to the relevant node, evict existing radio access network tasks from the node, and move those existing radio access network tasks to other nodes without causing disruptions to service.

Furthermore, the natural language processing module and the large language model can automatically coordinate with a resource management interface of the computing system to execute each of the commands in the operation sequence. As such, the disclosed system can perform the tasks defined by the natural language input with minimal manual intervention. In this way, the disclosed techniques can reduce the technical burden of operating a radio access network thereby streamlining day-to-day operations and improving overall efficiency.

In another technical benefit of the present disclosure, automating resource management tasks through a natural language interface and a large language model can improve service quality of the radio access network. As mentioned above, the complexity and custom development involved in operating a radio access network can result in errors and other technical difficulties. By automating resource management tasks through a natural language input, the disclosed system can minimize the risk of errors and potential service outages. For instance, a radio access network may require a custom endpoint for allocating computing resources, initializing workloads, and so forth. Such custom endpoints require compliance with specific protocols (e.g., E2/O1 protocols). Rather than place the technical burden of adhering to these protocols on the end user, the large language model can automatically generate the custom endpoint and translate the natural language input into the correct protocol specifications. In this way, the disclosed system can ensure custom configurations are consistently free from errors.

Features and technical benefits other than those explicitly described above will be apparent from a reading of the following Detailed Description and a review of the associated drawings. This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. The term “techniques,” for instance, may refer to system(s), method(s), computer-readable instructions, module(s), algorithms, hardware logic, and/or operation(s) as permitted by the context described above and throughout the document.

The techniques discussed herein enhance the functionality of computing systems that execute radio access networks providing a telecommunications service through a natural language interface and large language model to automate computing resource management tasks. As discussed above, managing a cloud radio access network can involve heavy manual customization of resource orchestration tools such as Kubernetes. Consequently, many traditional approaches can be error prone and unduly complex. In contrast, the features of the present disclosure enable computing systems to automate cloud radio access network management tasks to alleviate the technical complexity and reduce the risk of errors.

1 7 FIGS.- Various examples, scenarios, and aspects that enable natural language-based management of computing resources executing radio access network workloads are described below with respect to.

1 FIG. 100 102 104 104 102 102 102 102 104 106 106 104 104 104 102 illustrates a systemin which a computing environmentis controlled via a natural language input. In various examples, the natural language inputcan be provided by an end user (e.g., a technician, a system engineer) via a computing device that is connected to the computing environment. In various examples, the computing environmentcan include any computing infrastructure that executes radio access network functionalities. For instance, the computing environmentcan include a virtualized centralized unit (vCU) and a mobile core executed by a central computing system (e.g., a cloud datacenter) and a virtualized distributed unit (vDU) executed by an edge computing system. That is, the computing environmentcan comprise cloud computing devices, edge computing devices, and/or a combination of both cloud and edge computing devices. The natural language inputcan accordingly define a desired outcome. That is, the desired outcomecan be understood as the stated goal of the natural language input. In a specific example, the natural language inputcan be “scale out computing resources if CPU usage exceeds 95%”. As such, the desired outcome of the natural language inputcan be to firstly, monitor the computing resource usage of the computing environmentand secondly, allocate additional computing resources in the event the computing resource usage exceeds a predetermined threshold.

104 102 108 104 106 108 110 106 104 104 110 To process the natural language input, the computing environmentcan utilize a natural language process moduleto analyze the natural language inputand detect the desired outcome. The natural language processing modulecan accordingly retrieve an auxiliary information setbased on the desired outcomedefined by the natural language input. For example, continuing with the natural language inputmentioned above (“scale out computing resources if CPU usage exceeds 95%”), the auxiliary information setcan include documentation pertaining to monitoring and allocating computing resources for radio access network workloads.

108 112 104 110 112 The natural language processing modulecan subsequently generate a custom contextual querybased on the natural language inputand the auxiliary information set. In various examples, the custom contextual querycan also be referred to as a prompt. Generally described, a prompt commands a large language model to perform a certain task, typically in a natural language format (e.g., English). As advancements in large language models continue, prompt engineering has emerged as an important area of optimization to improve the performance of large language models. Depending on the format of the prompt, the large language model may produce vastly different outputs.

108 104 112 114 114 106 104 110 114 102 108 114 Consider one example, in which the natural language processing modulemerely provides the natural language input(“scale out computing resources if CPU usage exceeds 95%”) as the custom contextual queryto a large language model. Without additional context, information, or constraints, the large language modelmay fail to achieve the desired outcome. Now consider another example, in which the natural language inputis augmented by the auxiliary information setto provide the large language modelwith concrete examples and processes that are specific to the computing environment. In this way, the natural language processing modulecan constrain the behavior of the large language modeland ensure consistent performance.

108 112 114 114 116 118 106 110 118 110 118 114 116 116 114 114 Subsequently, the natural language processing modulecan provide the custom contextual queryto the large language model. In response, the large language modelcan generate an operation sequencecomprising a plurality of commandsthat perform one or more tasks to achieve the desired outcomein accordance with processes, protocols, and other aspects of the auxiliary information set. In a specific example, the commandscan be Kubernetes commands such as GET, PUT, PATCH, POST, and DELETE. For instance, the auxiliary information setcan include documentation outlining a process and/or tutorial on allocating additional computing resources to a specific radio access network workload utilizing various commands. Accordingly, the large language modelcan generate an operation sequencethat performs the process outlined by the documentation. In this way, the operation sequencegenerated by the large language modelcan streamline resource management operations. In various examples, the process defined by the documentation can be technically involved and thus time consuming if performed manually. In a specific example, migrating a radio access network workload to perform a software update without causing service disruptions can involve a complex series of commands that require significant technical expertise. As such, the large language modelcan alleviate the technical burden of end users such as system engineers and technicians thereby improving system efficiency.

116 120 120 120 114 122 124 126 128 The operation sequencecan then be utilized to configure a resource management interfacefor execution. Generally described, the resource management interfacecan be a software component that exposes the functionality of a resource management and orchestration system such as Kubernetes. In Kubernetes, the resource management interfacecan be analogous to an application programming interface (API) server that enables external components (e.g., the large language model) to communicate with the core components of a resource management and orchestration system such as a standard schedulerand a standard controller, as well as custom components such as a custom controllerwhich can operate on custom resource definitions.

128 102 130 132 120 132 130 134 120 128 134 120 128 126 124 120 128 134 In various examples, a custom resource definitioncan be a software component that is used to define a custom workload. For example, the computing environmentcan comprise a set of computing resources(e.g., CPUs, GPUs, virtual machines, containers) that execute various computing tasks. In a standard web application context, the resource management interfacecan assign standard computing tasks(i.e., workloads) to the computing resourcessuch as a Kubernetes deployment. However, for specialized workloads such as a radio access network, the resource management interfacemay require custom resource definitionsto support the specific functionalities of the radio access network. For example, the resource management interfacecan be configured with a “RAN” custom resource definition. Likewise, the custom controllercan be a specialized form of the standard controllerthat can communicate with the resource management interfaceand operate on custom resource definitionssuch as the radio access network.

124 102 128 120 130 132 134 126 130 120 134 126 In various examples, the custom controllercan monitor a state of the computing environment(e.g., compute and networking), some subset of a standard resource definition and/or the custom resource definitionsin the resource management interface, and the actual state of the computing resourceson the cluster. For example, the actual state of a standard computing resource can specify whether a computing task(e.g., a Kubernetes pod) has terminated. In another example, the state of the cluster can specify that a node needs to be upgraded. In still another example, the specification of the custom resource definition can specify a configuration for the radio access network. In response to changes to the state of the items being monitored, the custom controllercan in turn create, manage, and delete instances of the computing resources(e.g., Kubernetes pods, virtual machines) by communicating with the resource management interface. Due to the uniquely technical challenge of managing a radio access network, the actions performed by the custom controllercan be much more complex than simply creating a new Kubernetes pod on a cluster. In a specific example, moving a pod executing a virtualized distributed unit (vDU) in response to node migration can involve a complex series of timed tasks that require precise timing. In order to migrate a node, a new vDU pod is created, followed by insertion of a software switch container to replicate networking traffic, followed by communication with a pod executing a virtualized centralized unit (vCU) to migrate each piece of user equipment. After all the migrations are complete, the software switch and the old vDU pod can be removed.

118 116 120 122 124 126 128 130 132 106 104 104 106 130 108 104 110 108 110 Accordingly, the commandsof the operation sequencecan utilize the resource management interfaceand any of the standard scheduler, standard controller, custom controller, and custom resource definitionsto cause the computing resourcesto execute an automated computing taskto achieve the desired outcomedefined by the natural language input. In a specific example, the natural language inputcan state that “node A needs to be upgraded” in which the desired outcomeis applying an upgrade to “node A” of the computing resources. As such, the natural language processing modulecan analyze the natural language inputand retrieve an auxiliary information setthat pertains to upgrading nodes in a cloud radio access network context. For instance, the natural language processing modulecan have access to an internal database storing documentation defining various processes and best practices which can serve as the auxiliary information set.

104 110 108 112 104 110 114 102 134 112 114 116 118 120 132 132 132 134 Utilizing the natural language inputand the auxiliary information set, the natural language processing modulecan generate a custom contextual querythat incorporates the natural language inputand the auxiliary information set. In this way, the behavior of the large language modelcan be constrained within the specific context of the computing environmentexecuting the radio access network. Accordingly, the custom contextual querycan be provided to the large language modelwhich can generate an operation sequencecomprising a series of commandsthat cause the resource management interfaceto prevent computing tasksfrom being scheduled to the node (e.g., node A), evict existing computing tasksfrom the node, and move the existing computing tasksto another node utilizing various techniques to migrate the radio access networkwithout disruptions to service. As mentioned above, cloud radio access networks can be particularly sensitive to changes in state such as service outages and thus require specialized techniques for resource management and live migration. In this way, the disclosed techniques can maintain a consistent state of service and thus ensure service quality.

104 100 108 110 134 110 126 128 130 108 112 114 114 116 106 104 110 116 120 126 128 134 134 130 116 132 134 114 110 100 In another example, the natural language inputcan instruct the systemto “deploy a new radio access network”. Accordingly, the natural language processing modulecan retrieve an auxiliary information setpertaining to configuring and deploying a new radio access network. For instance, the auxiliary information setcan include documentation on generating the custom controllerand the custom resource definition, allocating computing resources, and proper configuration procedures. The natural language processing modulecan thusly generate a custom contextual queryto configure the large language model. In turn the large language modelcan generate an operation sequenceto carry out the desired outcomedefined by the natural language inputin accordance with the procedures defined in the auxiliary information set. For instance, the operation sequencecan, via the resource management interface, automatically generate requisite custom components such as the custom controllerand the custom resource definitionsfor a new radio access networkas well as components for configurating the new radio access networksuch as a YAML manifest which can be submitted to the computing resources. Moreover, the operation sequencecan configure a computing task(e.g., a custom workload) to serve as a custom endpoint for configuring the radio access network. By utilizing the large language modeland a well-defined knowledge set such the auxiliary information set, the systemcan ensure consistent service through automated resource management operations and reduce the risk of errors thereby improving overall efficiency.

2 FIG. 1 FIG. 200 202 204 102 202 102 102 Turning now to, aspects of a systemin which a computing environmentutilizes deeper integration with a large language modelto simplify system architecture are shown and described. As with the computing environmentdiscussed with respect to, the computing environmentcan include any computing infrastructure that executes radio access network functionalities. For instance, the computing environmentcan include a virtualized centralized component (vCU) and a mobile core executed by a central computing system (e.g., a cloud datacenter) and a virtualized distributed unit (vDU) executed by an edge computing system. That is, the computing environmentcan comprise cloud computing devices, edge computing devices, and/or a combination of both cloud and edge computing devices.

1 FIG. 102 114 104 116 134 114 116 126 128 134 114 As discussed above with respect to, a computing environmentcan integrate a large language modelto translate natural language inputsinto an operation sequenceto perform automated computing resource management tasks including specialized workloads such as the radio access network. Accordingly, the large language modelcan, via the operation sequence, automatically generate custom components such as the custom controllerand custom resource definitionsto support various functionalities of the radio access network. In this way, the large language modelcan alleviate the technical burden faced by end users (e.g., technicians, system engineers).

200 200 204 204 200 206 208 206 210 212 214 2 FIG. However, in some examples, through deeper integration of large language model functionality, the disclosed techniques can simplify system architectures by obviating custom components such as those illustrated in the systemshown in. As discussed above, manually managing custom workloads for a cloud radio access network can be a deep technical challenge often necessitating heavy customization of existing orchestration tools such as Kubernetes. In contrast, the systemcan utilize a large language modelto directly fulfill the functions of custom components. In this way, the large language modelcan enable efficient execution of complex tasks such as diagnosing performance issues and debugging. For instance, like the examples discussed above, the systemcan receive a natural language inputfrom an end user defining a desired outcome. The natural language inputcan be subsequently analyzed by a natural language processing moduleto retrieve a relevant auxiliary information setand generate a custom contextual query(e.g., a prompt).

214 204 216 218 202 208 206 204 218 120 130 134 120 126 120 134 204 1 FIG. Accordingly, the custom contextual querycan cause the large language modelto generate an operation sequencecomprising a plurality of commandsthat operate on the computing environmentto achieve the desired outcomedefined by the natural language input. However, unlike the examples discussed above with respect to, rather than utilize custom components, the large language modelcan generate commandsthat can directly communicate, via the resource management interface, to the computing resourcesexecuting a radio access network. In a specific example, consider a situation in the context of Kubernetes where the resource management interfacecan be the Kubernetes API server. Where some systems may require custom components (e.g., a custom controller) to act as intermediaries between the resource management interfaceand specialized workloads such as the radio access network, the large language modelcan obviate the custom components by instead utilizing standard Kubernetes commands to fulfill the functions of the custom components.

204 206 200 210 212 214 212 206 In a specific example, a simplified system architecture enabled by the large language modelcan streamline additional specialized workloads such as radio access network debugging. In this example, the natural language inputcan instruct the systemto “find the issue causing users to experience poor performance” where the desired outcome to diagnose an “issue causing users to experience poor performance”. In response, the natural language processing modulecan retrieve an auxiliary information setthat pertains to diagnosing performance issues such as documentation identifying troubleshooting processes. The resultant custom contextual querycan accordingly incorporate the auxiliary information setand the natural language input.

204 216 214 218 220 130 220 222 200 220 224 200 204 200 220 208 212 204 226 204 In turn, the large language modelcan generate an operation sequencein accordance with the custom contextual querycomprising a plurality of commandsto extract diagnostic datafrom the computing resourcesand/or the resource management interface. In various examples, the diagnostic datacan include log filesdefining events that took place at a certain time within the systemsuch as changed configurations, changed workloads, and so forth. In addition, the diagnostic datacan include metricsdefining various measures of performance of the systemsuch as resource utilization, latency, temperature, and the like. Leveraging the strong natural language processing capabilities of the large language model, the systemcan automatically analyze the diagnostic datain light of the desired outcomeand the auxiliary information setto detect specific issues. Accordingly, the analysis of the large language modelcan be provided to the user as a natural language output. In this way, the large language modelcan eliminate the need for a user to produce customized automation tools for diagnosing and debugging issues.

204 220 204 130 134 204 216 134 226 226 204 216 Furthermore, the large language modelcan suggest solutions to issues detected from analysis of the diagnostic data. For example, the large language modelcan determine that “the issue causing users to experience poor performance” is overloading of the computing resourcesexecuting the radio access network(e.g., heavy demand). In response, the large language modelcan generate an operation sequenceto allocate additional resources to the radio access networkto alleviate the performance issues. Alternatively, the natural language outputcan request user confirmation to allocate additional resources via the natural language output. In response to a user confirmation, the large language modelcan accordingly generate the operation sequence.

3 FIG. 300 302 304 304 306 306 304 306 Proceeding to, aspects of a systemfor enhancing a natural language processing modulefor generating a custom contextual query(e.g., a prompt) are shown and described. As mentioned above, prompt engineering has emerged as an important area of optimization in large language model performance. In various examples, the content of a custom contextual querycan dramatically affect the behavior and ultimately performance of a large language model. For example, a vague prompt that lacks context can cause the large language modelto behave erratically thereby leading to poor performance. In contrast, a custom contextual querythat includes detailed instructions as well as supplementary material to provide context can ensure consistent behavior and thus high performance from the large language model.

304 302 308 310 308 300 302 308 310 302 310 306 306 To generate a custom contextual query, the natural language processing modulecan primarily consider a user generated natural language inputthat defines a desired outcomewith respect to a radio access network. In a specific example, the natural language inputcan request the systemto “list the RAN instances on the cluster”. Subsequently, the natural language processing modulecan analyze the natural language inputto determine the desired outcome. In some examples, the natural language processing modulecan include an intrinsic instruction within the desired outcometo constrain the behavior of the large language model. For instance, the intrinsic instruction can constrain the large language modelwithin the context of a specific radio access network.

304 302 312 302 314 312 310 314 306 300 314 312 Likewise, to enhance the custom contextual query, the natural language processing modulecan be granted access to supplementary material such as a set of radio access network documentation. The natural language processing modulecan accordingly extract a specific subset of documentationfrom the radio access network documentationthat is most relevant to the desired outcome. In the present example, the documentationcan define what constitutes a “RAN instance” to provide context to the large language model. In this way, the systemcan limit the documentationto the most salient portions of the radio access network documentationthereby reducing processing times.

302 316 304 316 318 320 316 318 322 316 318 324 316 318 318 326 328 300 304 306 Furthermore, the natural language processing modulecan receive input from a pollerto further enhance the custom contextual query. In various examples, a pollercan be a software component that can automatically retrieve diagnostic datafrom a resource management interface. The pollercan be configured to retrieve the diagnostic dataat regular time interval(e.g., once per hour). In addition, the pollercan retrieve the diagnostic datain response to a trigger eventsuch as a resource failure. Moreover, the pollercan retrieve the diagnostic datain response to a user trigger. As discussed above, the diagnostic datacan include log filesand/or metricsto enable monitoring of a computing environment. In this way, the systemcan generate robust custom contextual queriesto ensure consistent performance from the large language model.

304 306 330 330 320 332 306 334 320 330 306 330 332 320 334 332 In accordance with the custom contextual query, the large language modelcan generate a series of commands(e.g., an API request) such as the operation sequences discussed above. The commandscan operate on computing resources executing a radio access network via the resource management interface. In addition, the commands can control various custom componentssuch as a custom controller and/or a custom resource definition as described in the examples above. Furthermore, the large language modelcan receive responsesfrom the resource management interfacein response to the individual commands. In a specific example, the large language modelcan generate a series of commandsto install a horizontal autoscaler custom componentto dynamically allocate additional resources to a radio access network. Accordingly, the resource management interfacecan generate a responseconfirming a successful installation of the custom component.

306 334 318 300 306 300 306 304 312 318 In various examples, the large language modelcan store a history of the responsesas well as diagnostic datato inform future behaviors and further adapt to the context of a cloud radio access network. Stated another way, the systemcan continue to train the large language modelin a live deployment context to improve performance over time. In this way, the systemcan ensure consistent behavior and high performance from the large language modelthrough robust custom contextual queriesthat are enhanced through domain specific radio access network documentationand diagnostic data.

4 FIG.A 400 402 404 402 404 402 406 404 402 404 404 Turning now to, aspects of a natural language user interfacefor interacting with a computing systemexecuting a radio access network are shown and described. As discussed above, an end user (e.g., a technician, a system engineer) can generate a natural language inputexpressing a desired outcome with respect to the computing systemand/or the radio access network therein. For example, the natural language inputA can request that the computing system“integrate support for multiple network interfaces in the cluster” to implement various network functionalities and be provided in a chat windowin which the user can freely generate natural language inputs. In response, as in the examples discussed above, the computing systemcan process the natural language inputA using a natural language processing module to generate a custom contextual query (e.g., a prompt) for a large language model. Subsequently, the large language model can generate an operation sequence to achieve the desired outcome defined by the natural language inputA.

402 408 404 408 404 408 404 408 404 Accordingly, the computing systemcan generate a first natural language outputA to acknowledge receipt of the natural language inputA. A second natural language outputB can confirm that the operation sequence fulfilling the request of the natural language outputA was successfully completed. In various examples, the first natural language outputA can be generated and displayed at a first time in response to receiving the natural language inputA. The second natural language outputB can be generated and displayed at a second time upon completion of the task initiated in response to the natural language inputA.

404 402 In another example, the user can provide a second natural language inputB requesting that the computing system“find the issue causing users to experience poor performance”. As discussed above, the large language model can generate an operation sequence to collect diagnostic data from the computing resources and/or the resource management interface. Moreover, the large language model can analyze the diagnostic data to determine the root cause of various issues such as resource utilization and resultant increases in latency.

402 408 402 408 410 402 400 400 Accordingly, the computing systemcan generate a third natural language outputC informing the user of the issue causing poor performance. As shown, the large language model of the computing systemcan determine that “RAN components are overloaded”. In addition, the third natural language outputC can suggest a solution to the problem identified by the large language mode from the diagnostic data. Namely, that the computing system can “allocate additional resources to address this issue”. In response to a confirmationfrom the user, the large language model of the computing systemcan proceed to allocate additional computing resources to resolve the issue. In this way, the natural language user interfacecan enable a user to implement new networking functionality within the radio access network with little technical burden. Moreover, the natural language user interfacecan enable the user to quickly identify and address issues in the radio access network with a likewise low technical burden.

4 FIG.B 4 FIG.A 412 406 414 416 416 416 402 Turning now to, an alternative natural language user interfaceis shown and described. While the chat windowdiscussed above with respect toenabled a user to freely produce natural language inputs (e.g., text, voice), the chat windowmay instead utilize a set of predetermined natural language inputs. In various examples, each of the predetermined natural language inputscan correspond to a respective predetermined query (e.g., a prompt). That is, selecting a predetermined natural language inputcan cause the computing systemto configure a large language model with the predetermined query.

416 418 402 402 420 418 402 422 418 418 422 422 418 402 406 414 Accordingly, the user can select from the set of predetermined natural language inputsto generate a first natural language inputA requesting that the computing system“diagnose an issue”. In response, the computing systemcan generate a natural language outputacknowledging the natural language inputA. In addition, the computing systemcan generate a second set of predetermined natural language inputsvia a large language model based on the context of the first natural language inputA. For instance, the first natural language inputA may indicate that an issue has occurred. As such, the second set of predetermined natural language inputscan relate to various potential issues within a radio access network. The user can subsequently select from the second set of predetermined natural language inputsto produce a second natural language inputB specifying the issue to be diagnosed. In this way, an operator can reduce the processing load on the computing systemthrough predetermined prompts and branching conversational dialogue. Consequently, an operator can elect to implement a free form chat windowas shown above and/or a branching chat windowbased on performance considerations, resource availability, and other factors.

5 FIG. 5 FIG. 500 502 Proceeding to, aspects of a routine 500 for translating a natural language input into an automated computing resource management task for a radio access network providing a telecommunications service are shown and described. With reference to, the routinebegins at operationwhere a natural language input defining a desired outcome for a computing system associated with the radio access network is received.

504 Next, at operation, the natural language input is analyzed utilizing the natural language processing function to detect the desired outcome for the computing system associated with the radio access network.

506 Then, at operation, an auxiliary information set pertaining to the radio access network is retrieved based on the desired outcome.

508 Subsequently, at operation, a custom contextual query is generated based on the desired outcome defined by the natural language input and the auxiliary information set.

510 Then, at operation, the custom contextual query is provided to a large language model. The custom contextual query causes the large language model to generate an operation sequence containing a plurality of commands defining the automated computing resource management task.

512 Finally, at operation, the computing system associated with the radio access network is configured with the operation sequence generated by the large language model based on the desired outcome defined by the natural language input and the auxiliary information set. When executed, the operation sequence causes the radio access network to perform the automated computing resource management task.

For ease of understanding, the process discussed in this disclosure are delineated as separate operations represented as independent blocks. However, these separately delineated operations should not be construed as necessarily order dependent in their performance. The order in which the process is described is not intended to be construed as a limitation, and any number of the described process blocks may be combined in any order to implement the process or an alternate process. Moreover, it is also possible that one or more of the provided operations is modified or omitted.

The particular implementation of the technologies disclosed herein is a matter of choice dependent on the performance and other requirements of a computing device. Accordingly, the logical operations described herein are referred to variously as states, operations, structural devices, acts, or modules. These states, operations, structural devices, acts, and modules can be implemented in hardware, software, firmware, in special-purpose digital logic, and any combination thereof. It should be appreciated that more or fewer operations can be performed than shown in the figures and described herein. These operations can also be performed in a different order than those described herein.

It also should be understood that the illustrated methods can end at any time and need not be performed in their entireties. Some or all operations of the methods, and/or substantially equivalent operations, can be performed by execution of computer-readable instructions included on a computer-storage media, as defined below. The term “computer-readable instructions,” and variants thereof, as used in the description and claims, is used expansively herein to include routines, applications, application modules, program modules, programs, components, data structures, algorithms, and the like. Computer-readable instructions can be implemented on various system configurations, including single-processor or multiprocessor systems, minicomputers, mainframe computers, personal computers, hand-held computing devices, microprocessor-based, programmable consumer electronics, combinations thereof, and the like.

Thus, it should be appreciated that the logical operations described herein are implemented (1) as a sequence of computer implemented acts or program modules running on a computing system and/or (2) as interconnected machine logic circuits or circuit modules within the computing system. The implementation is a matter of choice dependent on the performance and other requirements of the computing system. Accordingly, the logical operations described herein are referred to variously as states, operations, structural devices, acts, or modules. These operations, structural devices, acts, and modules may be implemented in software, in firmware, in special purpose digital logic, and any combination thereof.

500 For example, the operations of the routinecan be implemented, at least in part, by modules running the features disclosed herein can be a dynamically linked library (DLL), a statically linked library, functionality produced by an application programing interface (API), a compiled program, an interpreted program, a script, or any other executable set of instructions. Data can be stored in a data structure in one or more memory components. Data can be retrieved from the data structure by addressing links or references to the data structure.

500 500 Although the illustration may refer to the components of the figures, it should be appreciated that the operations of the routinemay be also implemented in other ways. In addition, one or more of the operations of the routinemay alternatively or additionally be implemented, at least in part, by a chipset working alone or in conjunction with other software modules. In the example described below, one or more modules of a computing system can receive and/or process the data disclosed herein. Any service, circuit, or application suitable for providing the techniques disclosed herein can be used in operations described herein.

6 FIG. 6 FIG. 600 100 600 602 604 606 608 610 604 602 602 602 602 602 602 shows additional details of an example computer architecturefor a device, such as a computer or a server configured as part of the cloud or edge platform or system, capable of executing computer instructions (e.g., a module or a program component described herein). The computer architectureillustrated inincludes processing system, a system memory, including a random-access memory(RAM) and a read-only memory (ROM), and a system busthat couples the memoryto the processing system. The processing systemcomprises processing unit(s). In various examples, the processing unit(s) of the processing systemare distributed. Stated another way, one processing unit of the processing systemmay be located in a first location (e.g., a rack within a datacenter) while another processing unit of the processing systemis located in a second location separate from the first location. For example, the processing systemcan include graphical processing units (GPUs) for executing complex artificial intelligence applications such as large language models. Moreover, the systems discussed herein can be provided as a distributed computing system such as a cloud/edge service.

602 Processing unit(s), such as processing unit(s) of processing system, can represent, for example, a CPU-type processing unit, a GPU-type processing unit, a field-programmable gate array (FPGA), another class of digital signal processor (DSP), or other hardware logic components that may, in some instances, be driven by a CPU. For example, illustrative types of hardware logic components that can be used include Application-Specific Integrated Circuits (ASICs), Application-Specific Standard Products (ASSPs), System-on-a-Chip Systems (SOCs), Complex Programmable Logic Devices (CPLDs), and the like.

600 608 600 612 614 616 618 A basic input/output system containing the basic routines that help to transfer information between elements within the computer architecture, such as during startup, is stored in the ROM. The computer architecturefurther includes a mass storage devicefor storing an operating system, application(s), modules, and other data described herein.

612 602 610 612 600 600 The mass storage deviceis connected to processing systemthrough a mass storage controller connected to the bus. The mass storage deviceand its associated computer-readable media provide non-volatile storage for the computer architecture. Although the description of computer-readable media contained herein refers to a mass storage device, the computer-readable media can be any available computer-readable storage media or communication media that can be accessed by the computer architecture.

Computer-readable media includes computer-readable storage media and/or communication media. Computer-readable storage media includes one or more of volatile memory, nonvolatile memory, and/or other persistent and/or auxiliary computer storage media, removable and non-removable computer storage media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. Thus, computer storage media includes tangible and/or physical forms of media included in a device and/or hardware component that is part of a device or external to a device, including RAM, static RAM (SRAM), dynamic RAM (DRAM), phase change memory (PCM), ROM, erasable programmable ROM (EPROM), electrically EPROM (EEPROM), flash memory, compact disc read-only memory (CD-ROM), digital versatile disks (DVDs), optical cards or other optical storage media, magnetic cassettes, magnetic tape, magnetic disk storage, magnetic cards or other magnetic storage devices or media, solid-state memory devices, storage arrays, network attached storage, storage area networks, hosted computer storage or any other storage memory, storage device, and/or storage medium that can be used to store and maintain information for access by a computing device.

In contrast to computer-readable storage media, communication media can embody computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave, or other transmission mechanism. As defined herein, computer storage media does not include communication media. That is, computer-readable storage media does not include communications media consisting solely of a modulated data signal, a carrier wave, or a propagated signal, per se.

600 620 600 620 622 610 600 624 624 According to various configurations, the computer architecturemay operate in a networked environment using logical connections to remote computers through the network. The computer architecturemay connect to the networkthrough a network interface unitconnected to the bus. The computer architecturealso may include an input/output controllerfor receiving and processing input from a number of other devices, including a keyboard, mouse, touch, or electronic stylus or pen. Similarly, the input/output controllermay provide output to a display screen, a printer, or other type of output device.

602 602 600 602 602 602 602 602 The software components described herein may, when loaded into the processing systemand executed, transform the processing systemand the overall computer architecturefrom a general-purpose computing system into a special-purpose computing system customized to facilitate the functionality presented herein. The processing systemmay be constructed from any number of transistors or other discrete circuit elements, which may individually or collectively assume any number of states. More specifically, the processing systemmay operate as a finite-state machine, in response to executable instructions contained within the software modules disclosed herein. These computer-executable instructions may transform the processing systemby specifying how the processing systemtransition between states, thereby transforming the transistors or other discrete hardware elements constituting the processing system.

7 FIG. 7 FIG. 700 700 700 depicts an illustrative distributed computing environmentcapable of executing the software components described herein. Thus, the distributed computing environmentillustrated incan be utilized to execute any aspects of the software components presented herein. For example, the distributed computing environmentcan be utilized to execute aspects of the software components described herein.

700 702 704 704 706 706 706 702 704 706 706 706 706 706 706 706 702 Accordingly, the distributed computing environmentcan include a computing environmentoperating on, in communication with, or as part of the network. The networkcan include various access networks. One or more client devicesA-N (hereinafter referred to collectively and/or generically as “computing devices”) can communicate with the computing environmentvia the network. In one illustrated configuration, the computing devicesinclude a computing deviceA such as a laptop computer, a desktop computer, or other computing device; a slate or tablet computing device (“tablet computing device”)B; a mobile computing deviceC such as a mobile telephone, a smart phone, or other mobile computing device; a server computerD; and/or other devicesN. It should be understood that any number of computing devicescan communicate with the computing environment.

702 708 610 712 708 708 714 716 718 720 722 708 724 7 FIG. In various examples, the computing environmentincludes servers, data storage, and one or more network interfaces. The serverscan host various services, virtual machines, portals, and/or other resources. In the illustrated configuration, the servershost virtual machines, Web portals, mailbox services, storage services, and/or social networking services. As shown inthe serversalso can host other services, applications, portals, and/or other resources (“other resources”).

702 710 710 704 710 700 710 726 726 726 726 808 726 726 As mentioned above, the computing environmentcan include the data storage. According to various implementations, the functionality of the data storageis provided by one or more databases operating on, or in communication with, the network. The functionality of the data storagealso can be provided by one or more servers configured to host data for the computing environment. The data storagecan include, host, or provide one or more real or virtual datastoresA-N (hereinafter referred to collectively and/or generically as “datastores”). The datastoresare configured to host data used or created by the serversand/or other data. That is, the datastoresalso can host or store web page documents, word documents, presentation documents, data structures, algorithms for execution by a recommendation engine, and/or other data utilized by any application program. Aspects of the datastoresmay be associated with a service for storing files.

702 712 712 712 The computing environmentcan communicate with, or be accessed by, the network interfaces. The network interfacescan include various types of network hardware and software for supporting communications between two or more computing devices including the computing devices and the servers. It should be appreciated that the network interfacesalso may be utilized to connect to other types of networks and/or computer systems.

700 700 700 It should be understood that the distributed computing environmentdescribed herein can provide any aspects of the software elements described herein with any number of virtual computing resources and/or other distributed computing functionality that can be configured to execute any aspects of the software components disclosed herein. According to various implementations of the concepts and technologies disclosed herein, the distributed computing environmentprovides the software functionality described herein as a service to the computing devices. It should be understood that the computing devices can include real or virtual machines including server computers, web servers, personal computers, mobile computing devices, smart phones, and/or other devices. As such, various configurations of the concepts and technologies disclosed herein enable any device configured to access the distributed computing environmentto utilize the functionality described herein for providing the techniques disclosed herein, among other aspects.

The disclosure presented herein also encompasses the subject matter set forth in the following clauses.

Example Clause A, a method for translating a natural language input into an automated computing resource management task for a radio access network providing a telecommunications service, the method performed by a computing environment associated with the radio access network, the method comprising: receiving the natural language input defining a desired outcome for the computing environment associated with the radio access network; analyzing the natural language input utilizing a natural language processing function to detect the desired outcome for the computing environment associated with the radio access network; retrieving, based on the desired outcome, an auxiliary information set pertaining to the radio access network; generating, by the natural language processing function, a custom contextual query based on the desired outcome defined by the natural language input and the auxiliary information set; providing the custom contextual query to a large language model of the natural language processing function, wherein the custom contextual query causes the large language model to generate an operation sequence containing a plurality of commands defining the automated computing resource management task; and configuring the computing environment associated with the radio access network with the operation sequence generated by the large language model based on the desired outcome defined by the natural language input and the auxiliary information set, wherein the operation sequence causes the computing environment associated with the radio access network to perform the automated computing resource management task.

Example Clause B, the method of Example Clause A, wherein the desired outcome defined by the natural language input comprises an implementation of networking functionality within the radio access network.

Example Clause C, the method of Example Clause B, wherein the operation sequence causes the computing system associated with the radio access network to retrieve and install a component facilitating the network functionality.

Example Clause D, the method of any one of Example Clause A through C, wherein the auxiliary information set comprises documentation that is specific to the computing system associated with the radio access network.

Example Clause E, the method of any one of Example Clause A through C, wherein the auxiliary information set comprises diagnostic information that is retrieved in response to an event within the radio access network.

Example Clause F, the method of any one of Example Clause A through E, wherein the custom contextual query includes an intrinsic instruction for constraining a behavior of the large language model.

Example Clause G, the method of any one of Example Clause A through F, wherein the natural language input is a predetermined natural language input that is selected from a set of predetermined natural language inputs.

Example Clause H, a system for translating a natural language input into an automated computing resource management task for a radio access network providing a telecommunications service, the system comprising: a processing system; and a computer readable medium having encoded thereon computer readable instructions that when executed by the processing system cause the system to perform operations comprising: receiving the natural language input defining a desired outcome for the radio access network; analyzing the natural language input utilizing a natural language processing function to detect the desired outcome for the radio access network; retrieving, based on the desired outcome, an auxiliary information set pertaining to the radio access network; generating, by the natural language processing function, a custom contextual query based on the desired outcome defined by the natural language input and the auxiliary information set; providing the custom contextual query to a large language model of the natural language processing function, wherein the custom contextual query causes the large language model to generate an operation sequence containing a plurality of commands defining the automated computing resource management task; and executing the operation sequence generated by the large language model based on the desired outcome defined by the natural language input and the auxiliary information set, wherein execution of the operation sequence causes the radio access network to perform the automated computing resource management task.

Example Clause I, the system of Example Clause H, wherein the desired outcome defined by the natural language input comprises an implementation of networking functionality within the radio access network.

Example Clause J, the system of Example Clause I, wherein the operation sequence causes the system to retrieve and install a component facilitating the network functionality.

Example Clause K, the system of any one of Example Clause H through J, wherein the auxiliary information set comprises documentation that is specific to the radio access network.

Example Clause L, the system of any one of Example Clause H through J, wherein the auxiliary information set comprises diagnostic information that is retrieved in response to an event within the radio access network.

Example Clause M, the system of any one of Example Clause H through J, wherein the custom contextual query includes an intrinsic instruction for constraining a behavior of the large language model.

Example Clause N, the system of any one of Example Clause H through M, wherein the natural language input is a predetermined natural language input that is selected from a set of predetermined natural language inputs.

Example Clause O, a computer readable storage medium having encoded thereon computer readable instructions that, when executed by a system, cause the system to perform operations comprising: receiving a natural language input defining a desired outcome for a radio access network; analyzing the natural language input utilizing a natural language processing function to detect the desired outcome for the radio access network; retrieving, based on the desired outcome, an auxiliary information set pertaining to the radio access network; generating, by the natural language processing function, a custom contextual query based on the desired outcome defined by the natural language input and the auxiliary information set; providing the custom contextual query to a large language model of the natural language processing function, wherein the custom contextual query causes the large language model to generate an operation sequence containing a plurality of commands defining an automated computing resource management task; and executing the operation sequence generated by the large language model based on the desired outcome defined by the natural language input and the auxiliary information set, wherein execution of the operation sequence causes the radio access network to perform the automated computing resource management task.

Example Clause P, the computer readable storage medium of Example Clause O, wherein the desired outcome defined by the natural language input comprises an implementation of networking functionality within the radio access network.

Example Clause Q, the computer readable storage medium of Example Clause P, wherein the operation sequence causes the system to retrieve and install a component facilitating the network functionality.

Example Clause R, the computer readable storage medium of any one of Example Clause O through P, wherein the auxiliary information set comprises documentation that is specific to the radio access network.

Example Clause S, the computer readable storage medium of any one of Example Clause O through P, wherein the auxiliary information set comprises diagnostic information that is retrieved in response to an event within the radio access network.

Example Clause T, the computer readable storage medium of any one of Example Clause O through S, wherein the custom contextual query includes an intrinsic instruction for constraining a behavior of the large language model.

Conditional language such as, among others, “can,” “could,” “might” or “may,” unless specifically stated otherwise, are understood within the context to present that certain examples include, while other examples do not include, certain features, elements and/or steps. Thus, such conditional language is not generally intended to imply that certain features, elements and/or steps are in any way required for one or more examples or that one or more examples necessarily include logic for deciding, with or without user input or prompting, whether certain features, elements and/or steps are included or are to be performed in any particular example. Conjunctive language such as the phrase “at least one of X, Y or Z,” unless specifically stated otherwise, is to be understood to present that an item, term, etc. may be either X, Y, or Z, or a combination thereof.

The terms “a,” “an,” “the” and similar referents used in the context of describing the invention (especially in the context of the following claims) are to be construed to cover both the singular and the plural unless otherwise indicated herein or clearly contradicted by context. The terms “based on,” “based upon,” and similar referents are to be construed as meaning “based at least in part” which includes being “based in part” and “based in whole” unless otherwise indicated or clearly contradicted by context.

In addition, any reference to “first,” “second,” etc. elements within the Summary and/or Detailed Description is not intended to and should not be construed to necessarily correspond to any reference of “first,” “second,” etc. elements of the claims. Rather, any use of “first” and “second” within the Summary, Detailed Description, and/or claims may be used to distinguish between two different instances of the same element (e.g., two different workloads)

In closing, although the various configurations have been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended representations is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as example forms of implementing the claimed subject matter.

In closing, although the various configurations have been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended representations is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as example forms of implementing the claimed subject matter.